复杂环境下单通道多信号盲分离与调制识别研究

With the increasing complexity environment of wireless communication, the signal, which has the property of large dynamic scall, dispersion, overlapping in time and frequency domain and impacted by non-gussian impulse interference, is received by communication system. For the complexity environment, the blind separation and modulation recognition in single-channel were main research object in the non-cooperative intelligent wireless communication system. Modern digital signal processing theory and combination-optimization theory were adopted. The methods of independent principle analysis, sparse decomposition, high order and fractional low order statistic signal processing and signal space decomposition were used comprehensively. To realize sparse representation of communication signals and then the algorithms for separating the blind source in single channel based on spectral filter and sparse decomposition, and estimating parameters and statistic parameters of communication signals’ cyclostationary statistical characteristics were set up. Then, the models based on the characteristic of linear periodic time variable and the modulation dictionaries based on the vectors of models’ parameters were constructed. Finally, the anti-noise performance, stability and separable degree would be analyzed and verified for the system of multi-channel modulated signals’ separating and recognizing. This project has theoretical significance and engineering application value in signal confirmation and information recovery in the fields of radio management, the Internet of things engineering and multiple communication system intercommunication.

随着无线通信环境的日益复杂，通信系统中接收到的信号呈现出动态范围大、弥散、时频域重叠且受到非高斯脉冲噪声干扰等特征。针对此环境，课题以非协作智能无线通信系统中单通道盲分离算法及调制方式识别方法为研究对象，采用现代数字信号处理理论和组合优化理论，综合利用独立分量分析、稀疏分解、高阶和分数低阶统计信号分析及信号空间分解等方法，实现通信信号的稀疏表示并建立通信信号低阶循环平稳统计特征及参数估计和基于谱域滤波器、稀疏分解的单通道盲源分离算法；构建基于通信信号线性周期时变（LPTV）模型和模型参数矢量所对应规律的调制识别字典；最后分析验证单通道多路调制信号分离与识别系统的抗噪性能、稳定度、可分离度及识别率。该课题的研究旨在无线电管理、物联网工程、多通信体制互通等领域信息确认与恢复具有理论意义和工程应用价值。

通信信号体制、调制方式的多样化以及通信场景、信道环境的复杂化，给数字调制识别和盲分离带了新的挑战。课题以非协作智能无线通信系统中单通道盲分离及调制方式识别为研究对象，采用稳定分布理论、调制信号信息结构和组合优化理论，综合利用独立分量分析、稀疏分解、高阶和分数低阶统计信号分析、谱估计及信号空间分解等方法，探究了复杂环境下单通道多信号的盲分离及通信信号识别理论、技术与方法。主要研究成果包括：.（1）提出了稳定分布下分数低阶时间平滑和频率平滑的快速算法，即FAM、SSCA、FSM算法，证明了稳定分布下通信信号的低阶循环谱与高斯下的循环谱结构具有一致性，为非高斯脉冲噪声环境下调制识别提供新的参考途径，并对通信信号载波频率、码元速率等参数进行估计。.（2）从信号自身的变化规律或特性以及信号产生的系统模型描述，研究了通信信号的时序结构及循环谱域特征，基于经验模式（EMD)模态分解、傅里叶基与小波基联合字典下信号空间分解、频移循环谱域及状态空间分析了单通道时频混叠多调制信号可分离机理及实现方法。综合利用同源信号的时序结构、变换域的变化相关性和低阶循环谱等多维结构信息，构造时谱测度代价函数，设计盲自适应滤波器实现多信号分离算法；并提出了基于状态空间的Kalman滤波、扩展Kalman粒子滤波(Extended Kalman Particle Filter, EPF)、无迹Kalman粒子滤波(Unscented Kalman Particle Filter, UPF)与循环神经网络(Recurrence Neural Network, RNN)的单通道盲源分离算法。.（3）研究了常规调制信号的时频域、循环谱、星座图、高阶累计量特征及线性周期时变（LPTV）模型。分析了非高斯噪声、多径衰减信道、单通道时频混叠多信号及大动态信噪比下的调制识别问题。基于高阶累积量、时域瞬时特征以及循环谱，构造了调制信号特征参量表，给出了MASK、MPSK、MFSK、MQAM类间识别和十二种信号类内识别方法以及Alpha稳定分布噪声下多载波OFDM信号与单载波2ASK, 4ASK, BPSK、4FSK, 16QAM信号的混合调制方式识别方法。